Github user httfighter commented on the issue:
https://github.com/apache/spark/pull/22683
@srowen @ajbozarth I am not sure about some things, can you give me some
advice? In the process of modification, I have a question. In Spark, whether M
and MB represent MiB. Spark does not use the unit of kilobytes to convert
between numbers.
private static final ImmutableMap<String, ByteUnit> byteSuffixes =
ImmutableMap.<String, ByteUnit>builder()
.put("b", ByteUnit.BYTE)
.put("k", ByteUnit.KiB)
.put("kb", ByteUnit.KiB)
.put("m", ByteUnit.MiB)
.put("mb", ByteUnit.MiB)
.put("g", ByteUnit.GiB)
.put("gb", ByteUnit.GiB)
.put("t", ByteUnit.TiB)
.put("tb", ByteUnit.TiB)
.put("p", ByteUnit.PiB)
.put("pb", ByteUnit.PiB)
.build();
<tr>
<td><code>spark.kryoserializer.buffer</code></td>
<td>64k</td>
<td>
Initial size of Kryo's serialization buffer, in KiB unless otherwise
specified.
Note that there will be one buffer <i>per core</i> on each worker. This
buffer will grow up to
<code>spark.kryoserializer.buffer.max</code> if needed.
</td>
</tr>
If this is the case, can we only guarantee the uniform use of 1024 for
digital conversion, no changes to the unit displays in log, UI, comments and
configured messages. Otherwise, we need to modify all the UI, log, comments,
and configuration information to ensure consistency, there is no guarantee that
all can be modified, and there will be no problems after the modification.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
